<![CDATA[This study compares the performance of the Generalized Autoregressive Conditional Heteroskedasticity (GARCH) and Long Short-Term Memory (LSTM) models in predicting daily volatility of the Jakarta Composite Index (JCI) for the 2016–2025 period. Volatility is an important indicator in assessing market risk and uncertainty, so accurate prediction methods are needed by investors, analysts, and policymakers. The JCI closing price data is converted into log returns and processed through cleaning, normalization, and sequence formation stages for modeling purposes. The GARCH(1,1) model is used to capture the nature of volatility clustering through a conditional variance approach, while LSTM is utilized to study non-linear patterns and long-term relationships in time series. The results show that GARCH(1,1) is able to describe volatility patterns in general, but is less responsive to sudden changes in volatility. In contrast, the LSTM model provides superior prediction performance with low prediction errors and high coefficient of determination values. These findings indicate that the deep learning approach is more effective in modeling the volatility dynamics of the Jakarta Composite Index (JCI) than traditional econometric methods, especially under volatile market conditions. Keywords: JCI Volatility, GARCH, LSTM, Time Series Forecasting, Deep Learning]]>
Copyrights © 2025
